Nice point, Shotglass.
its well known in literature as the anthropic principle of 3d vision
Ok, I wanna go back to the HDR thing for a little while now. The 3D thing is going nowhere...
One thing about HDR most seem to be missing is reflection. Let me try to visualize it for you;
First a short description of how reflections work. A lightsource sends out light, that light bounce off a surface hitting you eye, but some of it is absorbed and scattered. So the intensity of the reflection drops. Some surfaces, like mirrors, absorb very little, others, like charcoal, absorb a lot.
These two extremes are easy to simulate in any renderer.
Let's say you have an 8-bit renderer, giving you 256 shades of brightness. Also, let's ignore renderer output range for now.
You have a representation of the sun through a window, reflecting off a flat surface, like Don's example scene. The sun is very bright, so the designers give it the max brighness the renderer would allow, 255. The rays hit the surface, which has a reflectivity of 10%. So the reflection has a brightness of 25.
All who work with computer graphics or imagery of any kind know that a gray surface with the brighness value of 25 is not very bright. yet, i the real world, an object with a reflectionvalue of 10% would look very shiny indeed.
Let's do the same example with a renderer capable of 10 times the depth, 2550 shades of gray.
The sun is still very bright, so it gets a value of 2550. The surface still reflect 10%, so the reflection will be a brightness of 255.
Now to the part where it gets a little tricky.
Most of us, if not all, only have 8 bit outputs. So we need some sort of normalization, or tone mapping. The way this works is like described earlier in this thread. the scene is analyzed. The brightest spot in the scene is given the value 255, the darkest is given a value of 0. An 8-bit renderer don't have to worry about this, as no value will aver be over 255. The 'HDR' renderer however, need to consider this. So, looking at our scene, the sun is off camera, and the reflection is (for argument's sake) the brightest spot of the image. Analysis say the reflection is just within bounds, so no mapping needs to be done, and the image is rendered.
The real issue here isn't displayed dynamic range, it's how much dynamic range is available in the entire scene. Now go back and take a look at eekie's example from some game or other. Pay attention to the sun glinting off the water, and the reflections in general. And that particular scene even has the sun in the FOV.
I wish I was at work, I'd render a few examples of this for easier visualization....
One thing about HDR most seem to be missing is reflection. Let me try to visualize it for you;
First a short description of how reflections work. A lightsource sends out light, that light bounce off a surface hitting you eye, but some of it is absorbed and scattered. So the intensity of the reflection drops. Some surfaces, like mirrors, absorb very little, others, like charcoal, absorb a lot.
These two extremes are easy to simulate in any renderer.
Let's say you have an 8-bit renderer, giving you 256 shades of brightness. Also, let's ignore renderer output range for now.
You have a representation of the sun through a window, reflecting off a flat surface, like Don's example scene. The sun is very bright, so the designers give it the max brighness the renderer would allow, 255. The rays hit the surface, which has a reflectivity of 10%. So the reflection has a brightness of 25.
All who work with computer graphics or imagery of any kind know that a gray surface with the brighness value of 25 is not very bright. yet, i the real world, an object with a reflectionvalue of 10% would look very shiny indeed.
Let's do the same example with a renderer capable of 10 times the depth, 2550 shades of gray.
The sun is still very bright, so it gets a value of 2550. The surface still reflect 10%, so the reflection will be a brightness of 255.
Now to the part where it gets a little tricky.
Most of us, if not all, only have 8 bit outputs. So we need some sort of normalization, or tone mapping. The way this works is like described earlier in this thread. the scene is analyzed. The brightest spot in the scene is given the value 255, the darkest is given a value of 0. An 8-bit renderer don't have to worry about this, as no value will aver be over 255. The 'HDR' renderer however, need to consider this. So, looking at our scene, the sun is off camera, and the reflection is (for argument's sake) the brightest spot of the image. Analysis say the reflection is just within bounds, so no mapping needs to be done, and the image is rendered.
The real issue here isn't displayed dynamic range, it's how much dynamic range is available in the entire scene. Now go back and take a look at eekie's example from some game or other. Pay attention to the sun glinting off the water, and the reflections in general. And that particular scene even has the sun in the FOV.
I wish I was at work, I'd render a few examples of this for easier visualization....
That's because a) the virtual world is being recreated on the screen by a virtual camera and b) most game developers fancy themselves as film-makers
HDR works best in games like Counter Strike Source or Half life 2
I was just looking at the HDR in HL2 (screen shots of course)
and I noticed alot of upper contrast clipping, in lost coast you can't even see the clouds around the sun because its so clipped off.
And yet people say that looks realistic?
and I noticed alot of upper contrast clipping, in lost coast you can't even see the clouds around the sun because its so clipped off.
And yet people say that looks realistic?
Must Read! - HDR like effects for DX9 junkies!
(131 posts, started )
FGED GREDG RDFGDR GSFDG